CN108009285A - Forest Ecology man-machine interaction method based on natural language processing - Google Patents
Forest Ecology man-machine interaction method based on natural language processing Download PDFInfo
- Publication number
- CN108009285A CN108009285A CN201711408324.9A CN201711408324A CN108009285A CN 108009285 A CN108009285 A CN 108009285A CN 201711408324 A CN201711408324 A CN 201711408324A CN 108009285 A CN108009285 A CN 108009285A
- Authority
- CN
- China
- Prior art keywords
- natural language
- forest
- knowledge graph
- semantic
- knowledge
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Animal Behavior & Ethology (AREA)
- Human Computer Interaction (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Machine Translation (AREA)
Abstract
The present invention relates to a kind of Forest Ecology man-machine interaction method based on natural language processing, belong to field of neural networks.This method makes inferences the entity of the natural language in Forest Ecology using the method for knowledge graph, knowledge reasoning is set to be converted into by building the problem of deep neural network handles natural language question sentence, so as to find correspondence, represent the knowledge graph deep learning reasoning under study, obtain corresponding conclusion.The present invention introduces the concept of knowledge graph in deep learning, on the basis of knowledge graph is built, shallow semantic being understood, result injects knowledge graph, the semantic understanding of more deep layer is obtained by corresponding knowledge reasoning.The invention solves voice, text message interaction problems in Forest Ecology, multi-modal interactive device is realized the interaction functions such as forest zone navigation, localized weather consulting, ecological protection publicity, sight spot push, improves user experience effect.
Description
Technical field
The invention belongs to field of neural networks, is related to the Forest Ecology human-computer interaction side based on natural language processing
Method.
Background technology
As the key technology of intellectualization times, the artificial intelligence technology using deep learning as core has become a new round
The engine of Industrial Revolution, positive profound influence international industry competition situation and the international competitiveness of country.In point of artificial intelligence
In branch field, natural language processing (Natural Language Processing, NLP) is one therein representative neck
Domain, purpose are exactly computer is correctly handled human language, and make the various correct responses of people's expectation accordingly.Man-machine friendship
Mutual equipment is that one kind is complicated, integrated level is high, technology-intensive basic equipment, the development of new and high technology is led, in value
Chain is high-end and the key link of industrial chain.The development of many decades is undergone, human-computer interaction is presented on people in a manner of more and more natural
In face of.In recent years, the artificial intelligence technology of remarkable break-throughs and emerging virtual reality technology are obtained, deep reform man-machine
Interaction field, multi-modal, development trend that is strengthening across media, reality are substituting the interaction mode at traditional graph interface.
The content of the invention
In view of this, it is an object of the invention to provide a kind of man-machine friendship of the Forest Ecology based on natural language processing
Mutual method, the concept of knowledge graph is introduced in deep learning, on the basis of knowledge graph is built, shallow semantic is understood result
Knowledge graph is injected, the semantic understanding of more deep layer is obtained by corresponding knowledge reasoning.The invention is solved in Forest Ecology
Voice, text message interaction problems, make multi-modal interactive device realize forest zone navigation, localized weather consulting, ecological protection a surname
The interaction functions such as biography, sight spot push, improve user experience effect.
To reach above-mentioned purpose, the present invention provides following technical solution:
Forest Ecology man-machine interaction method based on natural language processing, comprises the following steps:
S1:Natural language text in the Forest Ecology of input is divided into word or phrase;
S2:Word is characterized as vector using word2vec, obtains matrix as input;
S3:Term vector merges, and the quantificational description of acquisition is fused into new term vector;
S4:Build the natural language deep learning model in Oriented Forest ecological environment;
S5:The natural language learning model optimization of Forest Ecology based on semi-supervised learning;
S6:Three-dimensional tensor knowledge graph structure under representative learning;
S7:Natural language knowledge graph deep learning reasoning in forest ecosystem under representative learning.
Further, the step S1 is specially:By the CRF parsers in HanLP and Stanford parser with
Text is divided into word or phrase by the interdependent parser of maximum entropy, and obtains part of speech, word order, keyword and dependence amount
Change description.
Further, the step S2 is specially:Utilize Word2vec (Word To Vector) neutral net language model
Term vector is trained, vocabulary is converted into vector form, thus the processing to text be converted into vector space to
Computing is measured, is readily achieved various NLP tasks.
Further, the step S3 is specially:According to the needs of different natural language processing tasks, term vector amalgamation mode
Take splicing, weighting or Hash calculation;Term vector syncretizing effect is by rarefaction representation process under the unsupervised learning that then carries out
Parameter carry out Comprehensive Evaluation.
Further, the step S4 is specially:By selecting suitable gradient to represent, intersection is carried out to learning model and is tested
Card, checks that learning model whether there is defect.
Further, the step S5 is specially:According to the basic representation structure of selection, by Forestry Ecological system under sparse representation
Natural language training data complete or collected works in system, the deep learning model for submitting to structure carry out unsupervised learning pre-training, obtain
Pre-training weights;After completing pre-training, to the partial data collection progress artificial knowledge for being no more than overall 20% in training data
These data are submitted to the deep learning model with pre-training weights according to same expression structure and have carried out prison by mark
Supervise and instruct white silk.
Further, the step S6 is specially:Known by defining triple (h, r, t) generation three-dimensional tensor to build semanteme
Know figure, wherein, h represents head semantic entity, and r represents semantic relation, and t represents tail semantic entity;By tensor resolution, one is obtained
A one factor matrix of core tensor sum, in core tensor each two-dimensional matrix section represent a kind of semantic relation, factor matrix
In represent a semantic entity per a line;The result reduced by core tensor sum factor matrix regards what corresponding triple was set up as
Probability.
Further, the step S7 is specially:Existing triple is extended using the template manually set, is generated
Natural language question sentence;Introduce word insertion concept by the knowledge graph training sample of acquisition be converted to for lower dimensional space vector, make knowledge
Reasoning is converted into by building the problem of deep neural network handles natural language question sentence, so that find " question sentence entity --- know
The correspondence of knowledge figure entity ", and the correspondence of " question sentence natural language description --- knowledge graph semantic relation ";Pass through
Hash, convolution, maximum pond and the Semantic mapping computing of the neural network model obtain answer type, answer path, answer afterwards
Three feature vectors of entity around case;These three feature vectors are done into similarity measure, final reasoning with question sentence vector respectively
Score is summed by three kinds of similarities and is obtained;That is S (q, a)=f1(q)Tg1(a)+f2(q)Tg2(a)+f3(q)Tg3(a), wherein, f1
(q)Tg1(a) similarity of answer type, f are represented2(q)Tg2(a) similarity in answer path, f are represented3(q)Tg3(a) represent to answer
The similarity of entity around case.
The beneficial effects of the present invention are:The present invention is using the method for knowledge graph to the natural language in Forest Ecology
Entity make inferences, make knowledge reasoning be converted into by build deep neural network handle natural language question sentence the problem of, from
And the correspondence of " question sentence entity -- knowledge graph entity " is found, and " question sentence natural language description --- knowledge graph is semantic to close
The correspondence of system ", represents the knowledge graph deep learning reasoning under study, obtains corresponding conclusion, make our natural language
Understand that function not only possesses the ability for understanding literal meaning, be also equipped with reasoning from logic, understand the ability of the deep layer meaning.
Brief description of the drawings
In order to make the purpose of the present invention, technical solution and beneficial effect clearer, the present invention provides drawings described below and carries out
Explanation:
Fig. 1 is Forest Ecology human-computer interaction of the present invention using the natural language processing of semi-supervised convolutional neural networks
The flow chart of technique construction;
Fig. 2 is the natural language understanding model in the Forestry Ecological under deep learning of the present invention;
Fig. 3 is the three-dimensional tensor knowledge graph schematic diagram under representative learning of the present invention;
Fig. 4 is the knowledge graph deep learning reasoning schematic diagram under present invention expression study;
Fig. 5 is that the semantic relation of knowledge entity of the present invention maps schematic diagram.
Embodiment
Below in conjunction with attached drawing, the preferred embodiment of the present invention is described in detail.
Fig. 1 is Forest Ecology human-computer interaction of the present invention using the natural language processing of semi-supervised convolutional neural networks
The flow chart of technique construction.Using the natural language text in forest ecosystem as semantic knowledge resource, knowledge graph is semantic table
Show method.A kind of natural language semantic knowledge figure based under deep neural network is built herein, utilizes the knowledge graph pair of structure
Entity in natural language is described.The semi-supervised convolutional neural networks of utilization are provided below in conjunction with the accompanying drawings to give birth to forestry
The embodiment of natural language semantic knowledge figure structure in state system is with the invention will be further elaborated.
As shown in Figure 1, each several part specific implementation details of the present invention are as follows:
1. the natural language text in the Forest Ecology of input is divided into word or phrase.By HanLP with
Text is divided into word or phrase by the CRF parsers in Stanford parser with the interdependent parser of maximum entropy, and
Obtain the quantificational descriptions such as part of speech, word order, keyword, dependence.
2. word is characterized as vector using word2vec, matrix is obtained as input.This mode of term vector is most important
Advantage is to allow the word for having certain relation, the distance in mathematical meaning closer to.To be trained to term vector, wherein
Most widely used method has neutral net language model, and word2vec is also based on what it was improved, herein for such a
Model grind making internal disorder or usurp.Word2vec (Word To Vector), vocabulary can be converted into vector form by it, so as to text
This processing is converted into the vector operation in vector space, is readily achieved various NLP tasks.Word2vec is with Forestry Ecological
Text in system builds a vocabulary first in training text data set, then trains each word as input
Term vector as output, the term vector file of generation can supply follow-up natural language processing and machine as feature vector
Learn scheduling algorithm to use.The position relationship of vocabulary in Word2Vec model extraction texts, extracts the contextual information of vocabulary, raw
Into the vector model of vocabulary.Vocabulary can represent that the similarity between vocabulary can be calculated by vector by numerical value vector quantization
Obtain.
3. term vector merges.The quantificational description of acquisition is fused into new term vector.Appointed according to different natural language processings
The needs of business, term vector amalgamation mode can take splicing, weighting or Hash calculation.Term vector syncretizing effect is by then carrying out
Unsupervised learning under rarefaction representation process parameter carry out Comprehensive Evaluation.
4. build the natural language deep learning model in the Oriented Forest ecosystem.By selecting suitable gradient table
Show, cross validation is carried out to learning model, check that learning model whether there is defect.
5. the natural language deep learning model optimization in the forest ecosystem under semi-supervised learning.Base according to selection
Natural language training data complete or collected works in forest ecosystem under sparse representation (are not carried out artificial knowledge's mark by this expression structure
Note), the deep learning model for submitting to structure carries out unsupervised learning pre-training, obtains pre-training weights.Wherein, set deep
Degree neutral net hidden layer is biased to 0, optimal value when being biased to assume weights ω=0 of output layer.Weights are arranged to ω ∈
(- r, r),Here faninFor preceding layer number of network node, fanoutFor latter layer network
Number of nodes.After completing pre-training, artificial knowledge's mark is carried out to the partial data collection (being no more than overall 20%) in training data
These data are submitted to the deep learning model with pre-training weights according to same expression structure and have carried out supervision by note
Training.
6. the three-dimensional tensor knowledge graph structure under representative learning.Come by defining triple (h, r, t) generation three-dimensional tensor
Semantic knowledge figure (as shown in Figure 3) is built, wherein, h represents head semantic entity, and r represents semantic relation, and it is semantic real that t represents tail
Body.By tensor resolution, one factor matrix of a core tensor sum is obtained, each two-dimensional matrix is cut into slices generation in core tensor
A kind of semantic relation of table, represents a semantic entity in factor matrix per a line.C is reduced by core tensor sum factor matrix
As a result the probability that corresponding triple is set up can be regarded as.
7. the natural language knowledge graph deep learning reasoning in forest ecosystem.Using the template manually set to existing
Triple be extended, generate natural language question sentence.Word insertion concept is introduced to be converted to the knowledge graph training sample of acquisition
For lower dimensional space vector, knowledge reasoning is set to be converted into by building the problem of deep neural network handles natural language question sentence, from
And the correspondence of " question sentence entity -- knowledge graph entity " is found, and " question sentence natural language description -- knowledge graph is semantic to close
The correspondence of system ", represents that the knowledge graph deep learning reasoning under study is as shown in Figure 4.Pass through the Kazakhstan of the neural network model
Uncommon, convolution, maximum pond and Semantic mapping computing obtain answer type, answer path, three kinds of features of entity around answer afterwards
Vector.These three feature vectors are done into similarity measure with question sentence vector respectively, final reasoning score is asked by three kinds of similarities
With and obtain.That is, S (q, a)=f1(q)Tg1(a)+f2(q)Tg2(a)+f3(q)Tg3(a).Wherein, f1(q)Tg1(a) answer class is represented
The similarity of type, f2(q)Tg2(a) similarity in answer path, f are represented3(q)Tg3(a) represent answer around entity it is similar
Degree.
Fig. 2 is the natural language understanding model in the Forestry Ecological under deep learning of the present invention.Base table according to selection
Show structure, the natural language training data complete or collected works (not carrying out artificial knowledge's mark) in Forestry Ecological under sparse representation submit
Deep learning model to structure carries out unsupervised learning pre-training, obtains pre-training weights.Wherein, depth nerve net is set
Network hidden layer is biased to 0, optimal value when being biased to assume weights ω=0 of output layer.Weights are arranged to ω ∈ (- r, r),Here faninFor preceding layer number of network node, fanoutFor later layer number of network node.
After completing pre-training, artificial knowledge's mark is carried out to the partial data collection (being no more than overall 20%) in training data, by this
A little data submit to the deep learning model with pre-training weights according to same expression structure and carry out Training.
Fig. 3 is the three-dimensional tensor knowledge graph schematic diagram under representative learning of the present invention.Generated by defining triple (h, r, t)
Three-dimensional tensor builds semantic knowledge figure (as shown in Figure 3), wherein, h represents head semantic entity, and r represents semantic relation, and t is represented
Tail semantic entity.By tensor resolution, one factor matrix of a core tensor sum is obtained, each Two-Dimensional Moment in core tensor
Battle array section represents a kind of semantic relation, and every a line represents a semantic entity in factor matrix.By core tensor sum factor matrix
The result of reduction can regard the probability that corresponding triple is set up as.
Fig. 4 is the knowledge graph deep learning reasoning schematic diagram under present invention expression study.Utilize the template pair manually set
Existing triple is extended, and generates natural language question sentence, and oneself in substantial amounts of Forestry Ecological is demarcated in a manner of Weakly supervised
Right speech training data, and obtain negative sample by the way of random disruptions have " question sentence -- answer " triple centering element.
Introduce word insertion concept by the knowledge graph training sample of acquisition be converted to for lower dimensional space vector, be converted into knowledge reasoning logical
The problem of crossing structure deep neural network processing natural language question sentence, so as to find pair of " question sentence entity -- knowledge graph entity "
It should be related to, and the correspondence of " question sentence natural language description -- knowledge graph semantic relation ".Know what word insertion study obtained
Know figure triple numerical value vector and use Recursivesentence basic representation structures, multiple row is submitted to after rarefaction expression
Convolutional neural networks model.By being obtained after the Hash of the neural network model, convolution, maximum pond and Semantic mapping computing
Three feature vectors of entity around to answer type, answer path, answer.These three feature vectors are vectorial with question sentence respectively
Similarity measure is done, final reasoning score is summed by three kinds of similarities and obtained.That is, S (q, a)=f1(q)Tg1(a)+f2(q)Tg2(a)+f3(q)Tg3(a).Wherein, f1(q)Tg1(a) similarity of answer type, f are represented2(q)Tg2(a) answer path is represented
Similarity, f3(q)Tg3(a) similarity of entity around answer is represented.
Fig. 5 is that the semantic relation of knowledge entity of the present invention maps schematic diagram.It is three-dimensional by defining triple (h, r, t) generation
Tensor builds semantic knowledge figure, wherein, h represents head semantic entity, r represents semantic relation, and t represents tail semantic entity.Assuming that
H and t is vectorial similar or equal with obtained by after the relevant mappings of r by certain, is defining energy function frUnder the premise of (h, t),
Build learning objective functionTo ensure the triple occurred in knowledge graph
Higher learning objective value is obtained, while the triple to not occurring is punished.Semantic relation is using map vector or mapping square
Battle array represents, by setting mapping function is mapped to the head entity of triple and tail entity and the relevant semanteme of relationship by objective (RBO)
The conversion of knowledge " 1-to-1 ", " 1-to-N ", " N-to-N " is realized in space.
Finally illustrate, preferred embodiment above is merely illustrative of the technical solution of the present invention and unrestricted, although logical
Cross above preferred embodiment the present invention is described in detail, however, those skilled in the art should understand that, can be
Various changes are made to it in form and in details, without departing from claims of the present invention limited range.
Claims (8)
1. the Forest Ecology man-machine interaction method based on natural language processing, it is characterised in that:This method includes following step
Suddenly:
S1:Natural language text in the Forest Ecology of input is divided into word or phrase;
S2:Word is characterized as vector using word2vec, obtains matrix as input;
S3:Term vector merges, and the quantificational description of acquisition is fused into new term vector;
S4:Build the natural language deep learning model in Oriented Forest ecological environment;
S5:The natural language learning model optimization of Forest Ecology based on semi-supervised learning;
S6:Three-dimensional tensor knowledge graph structure under representative learning;
S7:Natural language knowledge graph deep learning reasoning in forest ecosystem under representative learning.
2. the Forest Ecology man-machine interaction method according to claim 1 based on natural language processing, its feature exist
In:The step S1 is specially:Pass through the CRF parsers in HanLP and Stanfordparser and the interdependent sentence of maximum entropy
Text is divided into word or phrase by method analyzer, and obtains part of speech, word order, keyword and dependence quantificational description.
3. the Forest Ecology man-machine interaction method according to claim 1 based on natural language processing, its feature exist
In:The step S2 is specially:Term vector is instructed using Word2vec (WordToVector) neutral net language model
Practice, vocabulary is converted into vector form, so that the processing to text is converted into the vector operation in vector space, it is easily complete
Into various NLP tasks.
4. the Forest Ecology man-machine interaction method according to claim 1 based on natural language processing, its feature exist
In:The step S3 is specially:According to the needs of different natural language processing tasks, term vector amalgamation mode is taken splicing, is added
Power or Hash calculation;Term vector syncretizing effect is carried out comprehensive by the parameter of rarefaction representation process under the unsupervised learning that then carries out
Close and judge.
5. the Forest Ecology man-machine interaction method according to claim 1 based on natural language processing, its feature exist
In:The step S4 is specially:By selecting suitable gradient to represent, cross validation is carried out to learning model, checks study mould
Type whether there is defect.
6. the Forest Ecology man-machine interaction method according to claim 1 based on natural language processing, its feature exist
In:The step S5 is specially:According to the basic representation structure of selection, by the natural language in forest ecosystem under sparse representation
Say training data complete or collected works, the deep learning model for submitting to structure carries out unsupervised learning pre-training, obtains pre-training weights;It is complete
Into after pre-training, to the partial data collection progress artificial knowledge's mark for being no more than overall 20% in training data, these are counted
Training is carried out according to the deep learning model submitted to according to same expression structure with pre-training weights.
7. the Forest Ecology man-machine interaction method according to claim 1 based on natural language processing, its feature exist
In:The step S6 is specially:Semantic knowledge figure is built by defining triple (h, r, t) generation three-dimensional tensor, wherein, h
Head semantic entity is represented, r represents semantic relation, and t represents tail semantic entity;By tensor resolution, a core tensor sum is obtained
One factor matrix, in core tensor each two-dimensional matrix section represent a kind of semantic relation, represented in factor matrix per a line
One semantic entity;The result reduced by core tensor sum factor matrix regards the probability that corresponding triple is set up as.
8. the Forest Ecology man-machine interaction method according to claim 1 based on natural language processing, its feature exist
In:The step S7 is specially:Existing triple is extended using the template manually set, generation natural language is asked
Sentence;Introduce word insertion concept by the knowledge graph training sample of acquisition be converted to for lower dimensional space vector, be converted into knowledge reasoning
By building the problem of deep neural network handles natural language question sentence, so as to find " question sentence entity --- knowledge graph entity "
Correspondence, and the correspondence of " question sentence natural language description --- knowledge graph semantic relation ";Pass through the neutral net mould
The Hash of type, convolution, maximum pond and Semantic mapping computing obtain answer type, answer path, entity three around answer afterwards
Feature vectors;These three feature vectors are done into similarity measure with question sentence vector respectively, final reasoning score is by three kinds of phases
Obtained like degree summation;That is S (q, a)=f1(q)Tg1(a)+f2(q)Tg2(a)+f3(q)Tg3(a), wherein, f1(q)Tg1(a) represent to answer
The similarity of case type, f2(q)Tg2(a) similarity in answer path, f are represented3(q)Tg3(a) phase of entity around answer is represented
Like degree.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711408324.9A CN108009285B (en) | 2017-12-22 | 2017-12-22 | Forest Ecology man-machine interaction method based on natural language processing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711408324.9A CN108009285B (en) | 2017-12-22 | 2017-12-22 | Forest Ecology man-machine interaction method based on natural language processing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108009285A true CN108009285A (en) | 2018-05-08 |
CN108009285B CN108009285B (en) | 2019-04-26 |
Family
ID=62060716
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711408324.9A Active CN108009285B (en) | 2017-12-22 | 2017-12-22 | Forest Ecology man-machine interaction method based on natural language processing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108009285B (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108628834A (en) * | 2018-05-14 | 2018-10-09 | 国家计算机网络与信息安全管理中心 | A kind of word lists dendrography learning method based on syntax dependence |
CN110069632A (en) * | 2019-04-10 | 2019-07-30 | 华侨大学 | A kind of integrated shallow semantic indicates the deep learning file classification method of vector |
CN110309267A (en) * | 2019-07-08 | 2019-10-08 | 哈尔滨工业大学 | Semantic retrieving method and system based on pre-training model |
CN110599592A (en) * | 2019-09-12 | 2019-12-20 | 北京工商大学 | Three-dimensional indoor scene reconstruction method based on text |
CN111399629A (en) * | 2018-12-29 | 2020-07-10 | Tcl集团股份有限公司 | Operation guiding method of terminal equipment, terminal equipment and storage medium |
CN111626121A (en) * | 2020-04-24 | 2020-09-04 | 上海交通大学 | Complex event identification method and system based on multi-level interactive reasoning in video |
CN111832293A (en) * | 2020-06-24 | 2020-10-27 | 四川大学 | Entity and relation combined extraction method based on head entity prediction |
CN111862971A (en) * | 2020-06-15 | 2020-10-30 | 南京林业大学 | Forestry science popularization voice interaction method and control terminal system thereof |
CN112052685A (en) * | 2020-09-11 | 2020-12-08 | 河南合众伟奇云智科技有限公司 | End-to-end text entity relationship identification method based on two-dimensional time sequence network |
CN112348191A (en) * | 2020-10-26 | 2021-02-09 | 福州大学 | Knowledge base completion method based on multi-mode representation learning |
CN113590799A (en) * | 2021-08-16 | 2021-11-02 | 东南大学 | Weak supervision knowledge graph question-answering method based on multi-view reasoning |
CN113807940A (en) * | 2020-06-17 | 2021-12-17 | 马上消费金融股份有限公司 | Information processing and fraud identification method, device, equipment and storage medium |
US20210406467A1 (en) * | 2020-06-24 | 2021-12-30 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and apparatus for generating triple sample, electronic device and computer storage medium |
CN114417887A (en) * | 2022-03-22 | 2022-04-29 | 南京云问网络技术有限公司 | Natural language inference method and device fusing semantic parsing |
CN115033716A (en) * | 2022-08-10 | 2022-09-09 | 深圳市人马互动科技有限公司 | General self-learning system and self-learning method based on same |
CN116882502A (en) * | 2023-09-07 | 2023-10-13 | 山东山大鸥玛软件股份有限公司 | Professional text inference method and system integrating structured knowledge and text semantics |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106056207A (en) * | 2016-05-09 | 2016-10-26 | 武汉科技大学 | Natural language-based robot deep interacting and reasoning method and device |
US20170011289A1 (en) * | 2015-07-06 | 2017-01-12 | Microsoft Technology Licensing, Llc | Learning word embedding using morphological knowledge |
CN106547737A (en) * | 2016-10-25 | 2017-03-29 | 复旦大学 | Based on the sequence labelling method in the natural language processing of deep learning |
CN106855853A (en) * | 2016-12-28 | 2017-06-16 | 成都数联铭品科技有限公司 | Entity relation extraction system based on deep neural network |
CN107015963A (en) * | 2017-03-22 | 2017-08-04 | 重庆邮电大学 | Natural language semantic parsing system and method based on deep neural network |
US20170344880A1 (en) * | 2016-05-24 | 2017-11-30 | Cavium, Inc. | Systems and methods for vectorized fft for multi-dimensional convolution operations |
-
2017
- 2017-12-22 CN CN201711408324.9A patent/CN108009285B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170011289A1 (en) * | 2015-07-06 | 2017-01-12 | Microsoft Technology Licensing, Llc | Learning word embedding using morphological knowledge |
CN106056207A (en) * | 2016-05-09 | 2016-10-26 | 武汉科技大学 | Natural language-based robot deep interacting and reasoning method and device |
US20170344880A1 (en) * | 2016-05-24 | 2017-11-30 | Cavium, Inc. | Systems and methods for vectorized fft for multi-dimensional convolution operations |
CN106547737A (en) * | 2016-10-25 | 2017-03-29 | 复旦大学 | Based on the sequence labelling method in the natural language processing of deep learning |
CN106855853A (en) * | 2016-12-28 | 2017-06-16 | 成都数联铭品科技有限公司 | Entity relation extraction system based on deep neural network |
CN107015963A (en) * | 2017-03-22 | 2017-08-04 | 重庆邮电大学 | Natural language semantic parsing system and method based on deep neural network |
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108628834B (en) * | 2018-05-14 | 2022-04-15 | 国家计算机网络与信息安全管理中心 | Word expression learning method based on syntactic dependency relationship |
CN108628834A (en) * | 2018-05-14 | 2018-10-09 | 国家计算机网络与信息安全管理中心 | A kind of word lists dendrography learning method based on syntax dependence |
CN111399629B (en) * | 2018-12-29 | 2022-05-03 | Tcl科技集团股份有限公司 | Operation guiding method of terminal equipment, terminal equipment and storage medium |
CN111399629A (en) * | 2018-12-29 | 2020-07-10 | Tcl集团股份有限公司 | Operation guiding method of terminal equipment, terminal equipment and storage medium |
CN110069632B (en) * | 2019-04-10 | 2022-06-07 | 华侨大学 | Deep learning text classification method integrating shallow semantic expression vectors |
CN110069632A (en) * | 2019-04-10 | 2019-07-30 | 华侨大学 | A kind of integrated shallow semantic indicates the deep learning file classification method of vector |
CN110309267A (en) * | 2019-07-08 | 2019-10-08 | 哈尔滨工业大学 | Semantic retrieving method and system based on pre-training model |
CN110599592A (en) * | 2019-09-12 | 2019-12-20 | 北京工商大学 | Three-dimensional indoor scene reconstruction method based on text |
CN110599592B (en) * | 2019-09-12 | 2023-01-06 | 北京工商大学 | Three-dimensional indoor scene reconstruction method based on text |
CN111626121B (en) * | 2020-04-24 | 2022-12-20 | 上海交通大学 | Complex event identification method and system based on multi-level interactive reasoning in video |
CN111626121A (en) * | 2020-04-24 | 2020-09-04 | 上海交通大学 | Complex event identification method and system based on multi-level interactive reasoning in video |
CN111862971A (en) * | 2020-06-15 | 2020-10-30 | 南京林业大学 | Forestry science popularization voice interaction method and control terminal system thereof |
CN113807940B (en) * | 2020-06-17 | 2024-04-12 | 马上消费金融股份有限公司 | Information processing and fraud recognition method, device, equipment and storage medium |
CN113807940A (en) * | 2020-06-17 | 2021-12-17 | 马上消费金融股份有限公司 | Information processing and fraud identification method, device, equipment and storage medium |
CN111832293A (en) * | 2020-06-24 | 2020-10-27 | 四川大学 | Entity and relation combined extraction method based on head entity prediction |
US20210406467A1 (en) * | 2020-06-24 | 2021-12-30 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and apparatus for generating triple sample, electronic device and computer storage medium |
CN111832293B (en) * | 2020-06-24 | 2023-05-26 | 四川大学 | Entity and relation joint extraction method based on head entity prediction |
CN112052685A (en) * | 2020-09-11 | 2020-12-08 | 河南合众伟奇云智科技有限公司 | End-to-end text entity relationship identification method based on two-dimensional time sequence network |
CN112052685B (en) * | 2020-09-11 | 2024-06-04 | 河南合众伟奇云智科技有限公司 | End-to-end text entity relation identification method based on two-dimensional time sequence network |
CN112348191A (en) * | 2020-10-26 | 2021-02-09 | 福州大学 | Knowledge base completion method based on multi-mode representation learning |
CN113590799A (en) * | 2021-08-16 | 2021-11-02 | 东南大学 | Weak supervision knowledge graph question-answering method based on multi-view reasoning |
CN114417887A (en) * | 2022-03-22 | 2022-04-29 | 南京云问网络技术有限公司 | Natural language inference method and device fusing semantic parsing |
CN115033716A (en) * | 2022-08-10 | 2022-09-09 | 深圳市人马互动科技有限公司 | General self-learning system and self-learning method based on same |
CN115033716B (en) * | 2022-08-10 | 2023-01-20 | 深圳市人马互动科技有限公司 | General self-learning system and self-learning method based on same |
CN116882502A (en) * | 2023-09-07 | 2023-10-13 | 山东山大鸥玛软件股份有限公司 | Professional text inference method and system integrating structured knowledge and text semantics |
CN116882502B (en) * | 2023-09-07 | 2023-11-28 | 山东山大鸥玛软件股份有限公司 | Professional text inference method and system integrating structured knowledge and text semantics |
Also Published As
Publication number | Publication date |
---|---|
CN108009285B (en) | 2019-04-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108009285B (en) | Forest Ecology man-machine interaction method based on natural language processing | |
CN107578106B (en) | Neural network natural language reasoning method fusing word semantic knowledge | |
CN107239446B (en) | A kind of intelligence relationship extracting method based on neural network Yu attention mechanism | |
CN110825881B (en) | Method for establishing electric power knowledge graph | |
CN108133038A (en) | A kind of entity level emotional semantic classification system and method based on dynamic memory network | |
CN109558487A (en) | Document Classification Method based on the more attention networks of hierarchy | |
CN109271506A (en) | A kind of construction method of the field of power communication knowledge mapping question answering system based on deep learning | |
CN112131366A (en) | Method, device and storage medium for training text classification model and text classification | |
CN107662617B (en) | Vehicle-mounted interactive controlling algorithm based on deep learning | |
CN112860908A (en) | Knowledge graph automatic construction method based on multi-source heterogeneous power equipment data | |
CN106547735A (en) | The structure and using method of the dynamic word or word vector based on the context-aware of deep learning | |
CN110032635A (en) | One kind being based on the problem of depth characteristic fused neural network to matching process and device | |
CN106897254A (en) | A kind of network representation learning method | |
CN109933602A (en) | A kind of conversion method and device of natural language and structured query language | |
CN107423437A (en) | A kind of Question-Answering Model optimization method based on confrontation network intensified learning | |
CN110175229A (en) | A kind of method and system carrying out online training based on natural language | |
CN107679225A (en) | A kind of reply generation method based on keyword | |
CN113779220A (en) | Mongolian multi-hop question-answering method based on three-channel cognitive map and graph attention network | |
CN110347819A (en) | A kind of text snippet generation method based on positive negative sample dual training | |
CN112115242A (en) | Intelligent customer service question-answering system based on naive Bayes classification algorithm | |
Zhu et al. | Automatic short-answer grading via BERT-based deep neural networks | |
CN112364132A (en) | Similarity calculation model and system based on dependency syntax and method for building system | |
CN111598252B (en) | University computer basic knowledge problem solving method based on deep learning | |
CN116522945A (en) | Model and method for identifying named entities in food safety field | |
CN115600605A (en) | Method, system, equipment and storage medium for jointly extracting Chinese entity relationship |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |